arize-phoenix 11.10.0__py3-none-any.whl → 11.10.1__py3-none-any.whl

{arize_phoenix-11.10.0.dist-info → arize_phoenix-11.10.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: arize-phoenix
-Version: 11.10.0
+Version: 11.10.1
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://arize.com/docs/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues

{arize_phoenix-11.10.0.dist-info → arize_phoenix-11.10.1.dist-info}/RECORD

@@ -6,7 +6,7 @@ phoenix/exceptions.py,sha256=n2L2KKuecrdflB9MsCdAYCiSEvGJptIsfRkXMoJle7A,169
 phoenix/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
 phoenix/services.py,sha256=ngkyKGVatX3cO2WJdo2hKdaVKP-xJCMvqthvga6kJss,5196
 phoenix/settings.py,sha256=2kHfT3BNOVd4dAO1bq-syEQbHSG8oX2-7NhOwK2QREk,896
-phoenix/version.py,sha256=QWduDFvGZEp2yqC4pdaQmCm9ifKoDwiu-l0lM7_65A0,24
+phoenix/version.py,sha256=xvvOboOAs4Htuc0_lAWkZ6Ru9ZUbUhsIFFXYCIMHAAE,24
 phoenix/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/core/embedding_dimension.py,sha256=zKGbcvwOXgLf-yrJBpQyKtd-LEOPRKHnUToyAU8Owis,87
 phoenix/core/model.py,sha256=qBFraOtmwCCnWJltKNP18DDG0mULXigytlFsa6YOz6k,4837
@@ -176,8 +176,8 @@ phoenix/server/api/helpers/__init__.py,sha256=m2-xaSPqUiSs91k62JaRDjFNfl-1byxBfY
 phoenix/server/api/helpers/annotations.py,sha256=9gMXKpMTfWEChoSCnvdWYuyB0hlSnNOp-qUdar9Vono,262
 phoenix/server/api/helpers/dataset_helpers.py,sha256=3bdGBoUzqrtg-sr5p2wpQLOU6dhg_3TKFHNeJj8p0TU,9155
 phoenix/server/api/helpers/experiment_run_filters.py,sha256=DOnVwrmn39eAkk2mwuZP8kIcAnR5jrOgllEwWSjsw94,29893
-phoenix/server/api/helpers/playground_clients.py,sha256=vcLc8Ps8ZaZCUJnsN7retFbUR0LjF8nKTtN_OjBfS3E,71172
-phoenix/server/api/helpers/playground_registry.py,sha256=CPLMziFB2wmr-dfbx7VbzO2f8YIG_k5RftzvGXYGQ1w,2570
+phoenix/server/api/helpers/playground_clients.py,sha256=Fq4DNVIdnCiiVt0bh5mrZ7dJb2oOQcLjTttfq0Wcuv0,73589
+phoenix/server/api/helpers/playground_registry.py,sha256=n0v4-KnvZJxeaEwOla5qBbnOQjSWznKmMhZnh9ziJt0,2584
 phoenix/server/api/helpers/playground_spans.py,sha256=QpXwPl_fFNwm_iA1A77XApUyXMl1aDmonw8aXuNZ_4k,17132
 phoenix/server/api/helpers/prompts/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/server/api/helpers/prompts/models.py,sha256=nlPtLZaGcHfWNRR0iNRaBUv8eoKOnoGqRm6zadrTt0I,23547
@@ -437,9 +437,9 @@ phoenix/utilities/project.py,sha256=auVpARXkDb-JgeX5f2aStyFIkeKvGwN9l7qrFeJMVxI,
 phoenix/utilities/re.py,sha256=6YyUWIkv0zc2SigsxfOWIHzdpjKA_TZo2iqKq7zJKvw,2081
 phoenix/utilities/span_store.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/utilities/template_formatters.py,sha256=gh9PJD6WEGw7TEYXfSst1UR4pWWwmjxMLrDVQ_CkpkQ,2779
-arize_phoenix-11.10.0.dist-info/METADATA,sha256=-AwyriOi2Hypd_vYo7mrHEKmXIoypvFLTTmwn_MkdKA,30851
-arize_phoenix-11.10.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-arize_phoenix-11.10.0.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
-arize_phoenix-11.10.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
-arize_phoenix-11.10.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
-arize_phoenix-11.10.0.dist-info/RECORD,,
+arize_phoenix-11.10.1.dist-info/METADATA,sha256=wd_5sRdhTNXTvCYeob4xrZbZiPxZ81ENs-yF8PMEIyE,30851
+arize_phoenix-11.10.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+arize_phoenix-11.10.1.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
+arize_phoenix-11.10.1.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
+arize_phoenix-11.10.1.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
+arize_phoenix-11.10.1.dist-info/RECORD,,

phoenix/server/api/helpers/playground_clients.py

@@ -20,7 +20,7 @@ from openinference.semconv.trace import (
 )
 from strawberry import UNSET
 from strawberry.scalars import JSON as JSONScalarType
-from typing_extensions import TypeAlias, assert_never
+from typing_extensions import TypeAlias, assert_never, override
 
 from phoenix.config import getenv
 from phoenix.evals.models.rate_limiters import (
@@ -437,9 +437,9 @@ class OpenAIBaseStreamingClient(PlaygroundStreamingClient):
         if role is ChatCompletionMessageRole.TOOL:
             if tool_call_id is None:
                 raise ValueError("tool_call_id is required for tool messages")
-        return ChatCompletionToolMessageParam(
-            {"content": content, "role": "tool", "tool_call_id": tool_call_id}
-        )
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
         assert_never(role)
 
     def to_openai_tool_call_param(
@@ -1140,27 +1140,28 @@ class OpenAIStreamingClient(OpenAIBaseStreamingClient):
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
 
 
-@register_llm_client(
-    provider_key=GenerativeProviderKey.OPENAI,
-    model_names=[
-        "o1",
-        "o1-pro",
-        "o1-2024-12-17",
-        "o1-pro-2025-03-19",
-        "o1-mini",
-        "o1-mini-2024-09-12",
-        "o1-preview",
-        "o1-preview-2024-09-12",
-        "o3",
-        "o3-pro",
-        "o3-2025-04-16",
-        "o3-mini",
-        "o3-mini-2025-01-31",
-        "o4-mini",
-        "o4-mini-2025-04-16",
-    ],
-)
-class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
+_OPENAI_REASONING_MODELS = [
+    "o1",
+    "o1-pro",
+    "o1-2024-12-17",
+    "o1-pro-2025-03-19",
+    "o1-mini",
+    "o1-mini-2024-09-12",
+    "o1-preview",
+    "o1-preview-2024-09-12",
+    "o3",
+    "o3-pro",
+    "o3-2025-04-16",
+    "o3-mini",
+    "o3-mini-2025-01-31",
+    "o4-mini",
+    "o4-mini-2025-04-16",
+]
+
+
+class OpenAIReasoningReasoningModelsMixin:
+    """Mixin class for OpenAI-style reasoning model clients (o1, o3 series)."""
+
     @classmethod
     def supported_invocation_parameters(cls) -> list[InvocationParameter]:
         return [
@@ -1191,6 +1192,16 @@ class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
             ),
         ]
 
+
+@register_llm_client(
+    provider_key=GenerativeProviderKey.OPENAI,
+    model_names=_OPENAI_REASONING_MODELS,
+)
+class OpenAIReasoningNonStreamingClient(
+    OpenAIReasoningReasoningModelsMixin,
+    OpenAIStreamingClient,
+):
+    @override
     async def chat_completion_create(
         self,
         messages: list[
@@ -1283,46 +1294,11 @@ class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
         if role is ChatCompletionMessageRole.TOOL:
             if tool_call_id is None:
                 raise ValueError("tool_call_id is required for tool messages")
-        return ChatCompletionToolMessageParam(
-            {"content": content, "role": "tool", "tool_call_id": tool_call_id}
-        )
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
         assert_never(role)
 
-    @staticmethod
-    def _llm_token_counts(usage: "CompletionUsage") -> Iterator[tuple[str, Any]]:
-        yield LLM_TOKEN_COUNT_PROMPT, usage.prompt_tokens
-        yield LLM_TOKEN_COUNT_COMPLETION, usage.completion_tokens
-        yield LLM_TOKEN_COUNT_TOTAL, usage.total_tokens
-
-        if hasattr(usage, "prompt_tokens_details") and usage.prompt_tokens_details is not None:
-            prompt_details = usage.prompt_tokens_details
-            if (
-                hasattr(prompt_details, "cached_tokens")
-                and prompt_details.cached_tokens is not None
-            ):
-                yield LLM_TOKEN_COUNT_PROMPT_DETAILS_CACHE_READ, prompt_details.cached_tokens
-            if hasattr(prompt_details, "audio_tokens") and prompt_details.audio_tokens is not None:
-                yield LLM_TOKEN_COUNT_PROMPT_DETAILS_AUDIO, prompt_details.audio_tokens
-
-        if (
-            hasattr(usage, "completion_tokens_details")
-            and usage.completion_tokens_details is not None
-        ):
-            completion_details = usage.completion_tokens_details
-            if (
-                hasattr(completion_details, "reasoning_tokens")
-                and completion_details.reasoning_tokens is not None
-            ):
-                yield (
-                    LLM_TOKEN_COUNT_COMPLETION_DETAILS_REASONING,
-                    completion_details.reasoning_tokens,
-                )
-            if (
-                hasattr(completion_details, "audio_tokens")
-                and completion_details.audio_tokens is not None
-            ):
-                yield LLM_TOKEN_COUNT_COMPLETION_DETAILS_AUDIO, completion_details.audio_tokens
-
 
 @register_llm_client(
     provider_key=GenerativeProviderKey.AZURE_OPENAI,
@@ -1376,6 +1352,113 @@ class AzureOpenAIStreamingClient(OpenAIBaseStreamingClient):
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
 
 
+@register_llm_client(
+    provider_key=GenerativeProviderKey.AZURE_OPENAI,
+    model_names=_OPENAI_REASONING_MODELS,
+)
+class AzureOpenAIReasoningNonStreamingClient(
+    OpenAIReasoningReasoningModelsMixin,
+    AzureOpenAIStreamingClient,
+):
+    @override
+    async def chat_completion_create(
+        self,
+        messages: list[
+            tuple[ChatCompletionMessageRole, str, Optional[str], Optional[list[JSONScalarType]]]
+        ],
+        tools: list[JSONScalarType],
+        **invocation_parameters: Any,
+    ) -> AsyncIterator[ChatCompletionChunk]:
+        from openai import NOT_GIVEN
+
+        # Convert standard messages to OpenAI messages
+        openai_messages = []
+        for message in messages:
+            openai_message = self.to_openai_chat_completion_param(*message)
+            if openai_message is not None:
+                openai_messages.append(openai_message)
+
+        throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
+        response = await throttled_create(
+            messages=openai_messages,
+            model=self.model_name,
+            stream=False,
+            tools=tools or NOT_GIVEN,
+            **invocation_parameters,
+        )
+
+        if response.usage is not None:
+            self._attributes.update(dict(self._llm_token_counts(response.usage)))
+
+        choice = response.choices[0]
+        if choice.message.content:
+            yield TextChunk(content=choice.message.content)
+
+        if choice.message.tool_calls:
+            for tool_call in choice.message.tool_calls:
+                yield ToolCallChunk(
+                    id=tool_call.id,
+                    function=FunctionCallChunk(
+                        name=tool_call.function.name,
+                        arguments=tool_call.function.arguments,
+                    ),
+                )
+
+    def to_openai_chat_completion_param(
+        self,
+        role: ChatCompletionMessageRole,
+        content: JSONScalarType,
+        tool_call_id: Optional[str] = None,
+        tool_calls: Optional[list[JSONScalarType]] = None,
+    ) -> Optional["ChatCompletionMessageParam"]:
+        from openai.types.chat import (
+            ChatCompletionAssistantMessageParam,
+            ChatCompletionDeveloperMessageParam,
+            ChatCompletionToolMessageParam,
+            ChatCompletionUserMessageParam,
+        )
+
+        if role is ChatCompletionMessageRole.USER:
+            return ChatCompletionUserMessageParam(
+                {
+                    "content": content,
+                    "role": "user",
+                }
+            )
+        if role is ChatCompletionMessageRole.SYSTEM:
+            return ChatCompletionDeveloperMessageParam(
+                {
+                    "content": content,
+                    "role": "developer",
+                }
+            )
+        if role is ChatCompletionMessageRole.AI:
+            if tool_calls is None:
+                return ChatCompletionAssistantMessageParam(
+                    {
+                        "content": content,
+                        "role": "assistant",
+                    }
+                )
+            else:
+                return ChatCompletionAssistantMessageParam(
+                    {
+                        "content": content,
+                        "role": "assistant",
+                        "tool_calls": [
+                            self.to_openai_tool_call_param(tool_call) for tool_call in tool_calls
+                        ],
+                    }
+                )
+        if role is ChatCompletionMessageRole.TOOL:
+            if tool_call_id is None:
+                raise ValueError("tool_call_id is required for tool messages")
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
+        assert_never(role)
+
+
 @register_llm_client(
     provider_key=GenerativeProviderKey.ANTHROPIC,
     model_names=[

phoenix/server/api/helpers/playground_registry.py

@@ -1,4 +1,4 @@
-from typing import TYPE_CHECKING, Any, Callable, Optional, Union
+from typing import TYPE_CHECKING, Any, Callable, Optional, Sequence, Union
 
 from phoenix.server.api.types.GenerativeProvider import GenerativeProviderKey
 
@@ -59,7 +59,7 @@ PLAYGROUND_CLIENT_REGISTRY: PlaygroundClientRegistry = PlaygroundClientRegistry(
 
 def register_llm_client(
     provider_key: GenerativeProviderKey,
-    model_names: list[ModelName],
+    model_names: Sequence[ModelName],
 ) -> Callable[[type["PlaygroundStreamingClient"]], type["PlaygroundStreamingClient"]]:
     def decorator(cls: type["PlaygroundStreamingClient"]) -> type["PlaygroundStreamingClient"]:
         provider_registry = PLAYGROUND_CLIENT_REGISTRY._registry.setdefault(provider_key, {})

phoenix/version.py

@@ -1 +1 @@
-__version__ = "11.10.0"
+__version__ = "11.10.1"